Synthesis of Ni–Al Heterometallic Complexes

Combining equimolar amounts of [(BDI)AlH₂] (1; BDI = 2,6-diisopropylphenyl-β-methyldiketiminate) and [Ni(COD)₂] in toluene-d₈ at room temperature gave a gradual color change from yellow to red over the course of 1 h due to the formation of the heterometallic dihydride complex 2 (Scheme 1). ¹H NMR spectroscopy revealed an upfield signal at δ_H = −3.27 ppm assigned to the hydride ligands of 2 (c.f. δ_H  =  4.50 ppm for 1). This signal integrated for 2H with respect to a new (ⁱPr)-CH signal (4H) at δ_H = 3.15 ppm, suggesting that the new species had C_2v symmetry in solution. The reaction did not go to completion, and after 18 h, ¹H NMR spectroscopy showed both 2 along with starting materials, consistent with an equilibrium process. Addition of excess (10 equiv.) COD to mixtures containing 2 resulted in the reformation of 1 and [Ni(COD)₂]. A van‘t Hoff analysis was conducted to better understand and quantify the reversible process. The Gibbs free energy of binding was determined experimentally as ΔG°_exp = +2.9 kcal mol⁻¹ (Figures S25–S27) and calculated by density functional theory (DFT) as ΔG°_calc = +4.0 kcal mol⁻¹ (vide infra).

Acknowledging that the addition of a Lewis base might help to stabilize the product and thus push the reaction to completion by granting a five-coordinate aluminum center, 1 was combined with 1 equiv. of [Ni(COD)₂] and 4-dimethylaminopyridine (DMAP) in toluene-d₈ (Scheme 1). A dark red powder consistent with the formulation of 2·DMAP was isolated following workup in 74% yield. 2·DMAP is again in equilibrium with its starting materials, as addition of 10 equiv. COD to 2 reformed 1, [Ni(COD)₂] and DMAP (Figure S28). Based on ¹H NMR spectroscopic measurements, K_eq (2·DMAP) > K_eq (2) with the addition of DMAP appearing to favor product formation. This is likely due to the increased coordination number at aluminum leading to more hydridic [Al–H] units that have a greater affinity for nickel. In terms of diagnostic data, a new hydride signal of 2·DMAP appeared in the ¹H NMR spectrum with δ_H = ─4.59 ppm, integrating for 2H. The aromatic C─H signals for DMAP are found at δ_H = 8.88 and 5.94 ppm. These resonances are sharp and notably distinct from those of free DMAP (δ_H = 8.48 and 6.11 ppm). [Ni(COD)₂] appeared requisite to observe the five-coordinate aluminum center in 2·DMAP, as combining an equimolar mixture of DMAP and 1 in C₆D₆ yielded no apparent reaction at 298 K.^[30] Consistent with this observation, DMAP coordination to 1 was calculated to be endergonic with ΔG_calc = +3.6 kcal mol⁻¹, whereas addition of DMAP to 2 is exergonic with ΔG_calc = −3.7 kcal mol⁻¹ (Table S9).

Crystals of 2 were grown from a saturated n-pentane solution at −35 °C overnight. The solid-state structure of 2 (Figure 2a) features two tetrahedral metal centers (τ₄(Al) = 0.80; τ₄(Ni) = 0.84) and maintains a Ni—Al distance of 2.1953(10) Å, falling within the sum of covalent radii (∑_Ni,Al = 2.45 Å).^{[31, 32]} Such short metal–metal distances are not always indicative of a chemical bond (vide infra). The two bridging hydrides were located from the Fourier difference map; these bridge symmetrically and are 1.54(5) Å from Ni and 1.61(4) Å from Al. The crystallographic C₂ symmetry of the structure is further reflected in the planarity of the H─Ni─H′─Al core of 2. The average Al─H distances in 2 are around 0.1 Å longer than those in 1, but the errors and uncertainty of these measurements are such that definitive conclusions cannot be drawn. The solid-state structure of 2·DMAP was also confirmed by X-ray diffraction (Figure 2b). 2·DMAP now features a five-coordinate aluminum center with τ₅(Al) = 0.34, indicating a near-square pyramidal geometry. Upon DMAP coordination, the complex features a slight elongation of the Ni─Al (2.2606(14) Å) distance along with the introduction of asymmetry within the heterometallic core. Unlike 2, the Ni–(μ-H)₂–Al core in 2·DMAP is nonplanar, with the Ni atom showing the largest deviation of 0.17(4) Å from the mean plane.

To the best of our knowledge, 2 and 2·DMAP represent the first examples of structurally authenticated heterometallic Ni─Al complexes with two bridging hydride ligands. A Cambridge Structural Database search reveals only a handful of related structures that feature the Ni-(μ-H)-Al unit, with none containing more than one bridging hydride. For example, in 1990, Pörschke described a heterometallic complex consisting of a Me₂AlH moiety capped by 1-azabicyclo[2.2.2]octane, coordinated to [Ni(CDT)] (CDT = 1,5,9-cyclododecatriene) via a bridging hydride ligand.^{[12, 33-35]} This compound has a Ni─Al distance of 2.731(1) Å that is much longer than those found in 2 or 2⋅DMAP but has comparable Ni─H and Al─H distances.

Electronic Structures of 2 and 2·DMAP

The heterometallic complexes 2 and 2·DMAP were assessed by DFT calculations to elucidate the most appropriate bonding model.^{[36, 37]} Due to their high flexibility, careful conformational sampling was required to give accurate energies.^[38] In addition, Fractional Occupation number weighted electron Density (FOD) analysis revealed that the bimetallic complexes exhibited strong correlation effects (Figures S54–S63).^[39] After extensive benchmarking (Tables S3–S8), we report our results at the TPSS-D4/‌def2-QZVPP/SMD(toluene)//r²SCAN-3c/CPCM(toluene) level of theory. A series of canonical forms can be considered, including formulation as either Ni(0)/Al(III) or Ni(II)/Al(I) complexes depending on the nature of the metal–metal and metal–hydride bonding interactions (Figure 3a). Based on the analysis presented below, the most suitable bonding description of 2 and 2·DMAP is as bis(σ-alane) complexes of nickel with formal Ni(0)/Al(III) oxidation states.

Quantum theory of atoms in molecules (QTAIM) analysis for 2 and 2·DMAP showed no bond critical point (BCP) between Ni and Al, suggestive of limited metal–metal bonding. The presence of a BCP between each of the hydride ligands and metal centers indicated that the bonding in the bimetallic cores is dominated by bridging interactions through the hydrides (Figure 3b). These BCPs featured significant positive values for the Laplacian of the electron density (∇²ρ(r)), indicative of the largely electrostatic nature of the metal–hydride interactions (Table S13).

Natural bond orbital (NBO) analysis gave a qualitatively similar model, fragmenting 2 and 2·DMAP across the Ni─H bonds. Charges from natural population analysis (NPA) were positive on both Al (2, +1.42; 2·DMAP, +1.45) and Ni (2, +0.32; 2·DMAP, +0.33) and negative on the hydride ligands (2, −0.35/−0.36; 2·DMAP, −0.34/−0.35). The Wiberg bond indices (WBIs) showed that the bimetallic core features minimal Ni─Al bonding (2, 0.26; 2·DMAP, 0.28) along with 3c-2e interactions in which the hydrides are most strongly associated with the Al rather than the Ni nuclei (2, Al─H = 0.75/0.75, Ni─H = 0.14/0.14; 2·DMAP Al─H = 0.44/0.46, Ni─H = 0.18/0.18). Mayer bond indices suggested a drop in the Ni─Al (2, 0.90; 2·DMAP, 0.55) covalency upon coordination of DMAP to 2 to form 2·DMAP. This can be attributed to population of the Al 3p orbital by the Lewis base, resulting in decreased Ni 3d → Al 3p back-bonding (Table S11).

A more detailed MO description of the bonding could be obtained from ETS-NOCV analysis, which identified a total orbital contribution to the interaction of ΔE_orb = −73.2 kcal mol⁻¹ (Figures 3c and S64; Table S14). The bonding in 2 is dominated by donation of electron density from the two Al─H σ-bonds to the vacant Ni 3d_yz orbital, with significant back-donation from the filled Ni 3d_xy and 3d_z2 orbitals to a combination of s, p, and d acceptor orbitals on Al as well as the Al─H σ* antibonding orbitals. Agreeing with the Mayer population analysis, the coordination of DMAP results in consistently lower Al contributions to the analogous orbital interactions for 2·DMAP (Figure S65, Table S15).

C(sp²)─H Activation of DMAP and Pyridine

Given the newfound access to 2·DMAP, we wondered whether the coordination event might bring about the possibility of substrate C(sp²)─H bond activation. A C₆D₆ solution of 2·DMAP was monitored overnight at room temperature by ¹H NMR spectroscopy and showed partial consumption of 2·DMAP (Figure S18). During this process the aromatic signals of DMAP are desymmetrized, with diagnostic signals appearing for the product at δ_H = 7.39 and 5.82 ppm, each integrating for 1H along with a third resonance between δ_H = 7.17–7.21 overlapping with the β-diketiminate ligand environments. At the same time, the hydride signal of 2·DMAP at δ_H = −4.59 ppm was observed to disappear with the appearance of a new resonance at δ_H = −3.51 ppm. Collectively, these data support a room temperature C(sp²)─H activation process at the 2-position to give 3 (Figure 4a). 3 can be isolated in 61% yield and forms with the elimination of 1 equiv. of H₂.

The connectivity of 3 was confirmed by X-ray diffraction (Figure 4b). Complex 3 features a five-membered cyclic core with a hydride and heteroaryl ligand that bridge Ni and Al centers. The nitrogen of the now C(sp²)─H activated DMAP motif maintains coordination to aluminum. Comparatively, the Ni─Al distance of 2.3189(8) Å is longer than those in 2 or 2·DMAP. The Al─N distance to the pyridyl ligand is 1.882(2) Å in 3. This is shorter than that of 2.038(3) Å observed in 2·DMAP, consistent with conversion from a neutral to anionic ligand. At the same time, the C─N distance to the metalated 2-position of the ring becomes longer on C(sp²)─H activation.^[40]

The conversion of 2·DMAP to 3 occurs slowly with first-order kinetics at room temperature (k_obs = 1.3 x 10⁻³ min⁻¹). This reaction can, however, be accelerated by the addition of a catalytic quantity of an exogenous ligand. Addition of 20 mol% PCy₃ to 2·DMAP led to rapid conversion to 3 with a near fourfold increase in rate reaction (k_obs = 6.2 x 10⁻³ min⁻¹) and improved yields. The kinetics of this transformation were studied. Varying the concentration of PCy₃ revealed a linear relation (R² = 0.98) between k_obs and [PCy₃], indicating a first-order process. The catalytic effect of PCy₃ is irrefutable, as in its absence, mixtures of [Ni(COD)₂], DMAP, and 1 produce at maximum 20% of 3 after 240 h (Figure 4c) at 298 K (c.f. 55% of 3 using 20 mol% PCy₃ after 6 h, as measured by ¹H NMR spectroscopy). Despite its impact on the rate and efficiency of C(sp²)─H activation, there is no evidence for the formation of stable coordination complexes from the binding of PCy₃ to 2·DMAP or 3 at 298 K. The C─H activation of DMAP also appears to be reversible: addition of H₂ (1 bar) to a solution of 3 showed the reformation of the starting materials (Figure S24).

Attempts to expand this reactivity to an array of pyridine-based substrates provided variable outcomes. No reaction was observed between 1, [Ni(COD)₂], and 2,6-lutidine, either in the presence or absence of PCy₃ (Figure S22). Quinoline underwent direct hydroalumination and dearomatization with 1 catalyzed by 10 mol% [Ni(COD)₂]. Pyridine itself proved more successful. While combining 1 equiv. of [Ni(COD)₂], 1, and pyridine in C₆D₆ showed no signs of binding or C(sp²)─H activation by ¹H NMR spectroscopy, addition of PCy₃ (20 mol%) to this solution produced 4 in 78% conversion over 18 h (Figure S23). These results further support an important role of PCy₃ as a catalyst for C(sp²)─H bond activation in the absence of a strong para-electron donating group such as ─NMe₂.

To gain more insight into the mechanism of C(sp²)─H activation, the reaction was performed with pyridine-d₅. A kinetic isotope effect (KIE) of 1.1 ± 0.1 was found at 298 K through side-by-side kinetic runs (Figure 4d). A similar KIE of 0.9 ± 0.2 was determined through a competition experiment in which a mixture of 10 equiv. of pyridine and 10 equiv. of pyridine-d₅ was added to [Ni(COD)₂] and 1. These values indicate either no or a very small KIE, strongly suggesting that the rate-determining step is not the cleavage of the C(sp²)─H bond. For comparison, reactions involving oxidative addition of C(sp²)─H bonds to transition metals can occur with KIEs of up to 7–8.^[41] Activation of pyridine by an Fe–Al heterometallic complex was found to proceed with a very large primary KIE of 14.0 ± 0.2.^[10] Furthermore, when considering the fate of the isotopic label, during the reaction of pyridine-d₅, [Ni(COD)₂], and 1, no D-incorporation was observed in the bridging hydride position of the product 4, suggesting the isotopic label is lost exclusively through elimination of H–D.

The conversion of 2⋅DMAP to 3 provides a rare example of a defined reaction in which both C(sp²)─H activation and H₂ elimination are observed at a heterometallic complex. We thus set out to gain deeper insight into the mechanism of phosphine-catalyzed C(sp²)─H activation and H₂ elimination by computational means; for clarity, some intermediates in the steps that exchange COD to PCy₃ have been omitted but are all energetically accessible. (Figure 5 and S42).

The COD to PCy₃ ligand exchange process at 2·DMAP was calculated to occur through a stepwise pathway. It is initiated by decoordination of one of the η²-coordinated alkene arms of COD, transforming it from a bidentate to monodentate ligand in Int-1. This was calculated to be an uphill but thermodynamically accessible process (ΔG°_298K = +10.8 kcal mol⁻¹). Subsequent coordination of PCy₃ forms Int-2 (ΔG°_298 K = +13.6 kcal mol⁻¹) followed by the dissociation of COD, generating key intermediate Int-3 (ΔG°_298 K = +9.4 kcal mol⁻¹). C(sp²)─H activation proceeds from an active conformer of Int-3, Int-3′ (ΔG°_298 K = +17.5 kcal mol⁻¹), which is stabilized through an η²-C─H agostic interaction with Ni. This leads directly to the C─H bond-breaking transition state TS-1 (ΔG^‡_298K = +18.2 kcal mol⁻¹), yielding the oxidative addition product Int-4, which features a Ni─C(sp²) bond and a terminal hydride ligand on Ni, in addition to the two hydrides bridging Ni and Al centers. The reductive elimination of dihydrogen from Int-4 is a multistep process, initiated by migration of the bridging hydride ligand (TS-2, ΔG^‡_298 K = +13.5 kcal mol⁻¹) to a terminal position to form Int-5 (ΔG°_298 K = +12.4 kcal mol⁻¹). Reductive coupling of H₂ occurs through TS-3 (ΔG^‡_298K = +12.8 kcal mol⁻¹) with formation of the dihydrogen complex Int-6 (ΔG°_298 K = +10.4 kcal mol⁻¹). The dissociation of H₂ and ligand exchange of PCy₃ for COD completes the reaction sequence, regenerating the phosphine, the product 3, and a molecule of H₂ (ΔG°_298 K = −1.0 kcal mol⁻¹).

NBO calculations were performed to gain insight into the changes in electronic structure across the reaction pathway and the role of phosphine in catalyzing the C─H activation process. The C─H activation of DMAP gave more positive charge on the two metals as the reaction progressed, consistent with a formal oxidation from Ni(0)/Al(III) to Ni(II)/Al(III). Specifically for the series from Int-3 → Int-3′ → TS-1 → Int-4 the NPA charges on Ni trend −0.11 → +0.13 → +0.17 → +0.19, while those on Al trend +1.45 → +1.52 → +1.59, →+1.61.

To support the catalytic role of PCy₃, alternative C─H activation mechanisms ligated only by COD were also explored (Figures S43–S51). The most accessible phosphine-free transition state for the C─H activation of DMAP is a direct analogue of TS-1, supported by an η²-COD ligand (TS-1a, ΔG^‡_298 K = +30.2 kcal mol⁻¹). Comparing the electronic structure of the key transition states TS-1 to TS-1a suggests that PCy₃ likely accelerates the C─H activation step through increasing the electron density at Ni, resulting in increased population of the C─H σ* antibonding orbital of the pyridine fragment, allowing for bond breaking (NPA charge on Ni: TS-1: +0.17; TS-1a: +0.36).^[42] Alternatively, reactivity might be promoted through access to a three-coordinate Ni site in Int-3, which is expected to be more reactive than the four-coordinate center in 2·DMAP.

The DFT calculations suggest that the oxidative addition of the C(sp²)─H bond of DMAP to 1 is a low-energy process that, when catalyzed by PCy₃, is facile at room temperature. This observation is not only consistent with the lack of significant KIE in the reaction of 1 with pyridine catalyzed by PCy₃,^[41] but also with numerous studies involving two-component Ni/Al catalyst systems for the C–H functionalization of pyridines, which occur with low KIEs ranging between 1.04 and 1.25.^{[16, 19, 20]} These data suggest that C─H activation in these systems may be a universally low-energy step, one that is not rate limiting. Similarly, the DFT calculations suggest that reductive elimination of H₂ en route to form 3 proceeds through an almost flat potential energy surface and is also unlikely to be rate-limiting. This prediction is consistent with the experimental observation that no hydride scrambling between bridging and terminal positions occurs in the reaction of 1, cat. PCy₃, and pyridine-d₅. Rather, based on the lack of experimental KIE and 1st-order dependence of kinetics in PCy₃, we suggest that an activation barrier associated with the COD to PCy₃ ligand-exchange sequence is likely rate-determining.